NOTABLE QUOTES
“So it means there's 4% software, 96% service, which is bananas.”
Max Junestrand · 34:09 — Crystallizes the episode's legal-AI thesis — a trillion-dollar market that is almost entirely un-softwared.
“Which means to drag it out.”
Max Junestrand · 37:14 — Names the misaligned billable-hour incentive (lawyers paid to slow deals) that AI and founder-led diligence disrupt.
“You cannot build a legal research solution that doesn't have all of the data.”
Max Junestrand · 45:40 — The counterintuitive all-or-nothing data requirement that separates legal research from a typical power-law market.
“you hit the ceiling or you understand how shallow it is, and then you call us. So it's actually a big pipeline generator for us.”
Max Junestrand · 47:49 — Legora's framing of Anthropic's Claude legal product as a lead funnel rather than a competitive threat.
“Yeah, the voice is identity and IP.”
Mati Stanislawski · 18:23 — The one-line thesis behind ElevenLabs' cloning safeguards, detection tooling, and paid voice marketplace.
“we're kind of there on AGI, don't you think? That we've kind of achieved it, we just haven't deployed it?”
Jason Calacanis · 30:23 — Jason's provocative AGI framing capping the ElevenLabs segment on how far voice and general models have come.
PREDICTIONS ON THE RECORD
Dated and falsifiable — full record at /api/episode/<id>.
Jason Calacanis · unspecified · 29:12
ElevenLabs will end up building its own language model by forking the many strong open-source models now available, given its resources.
“It's pretty clear in my estimation that that's where you, you will wind up and the ability to make your own language model today, especially with all these great models out there that are now open-sourced, it's going to be pretty easy for a company with your level of resources.”
Mati Stanislawski · this year · 29:41
This year ElevenLabs will make any voice conversation feel indistinguishable from speaking with another human (a voice Turing test).
“And we hope this year we'll do the same thing for voice, where any conversation feels like you are speaking with another human.”
Mati Stanislawski · unspecified · 9:58
The customer-help interface will shift from reactive to proactive, with voice surfacing information and help before the user even asks.
“It will shift from reactive to proactive to help you get that help before you potentially ask for it.”
Max Junestrand · unspecified · 37:22
Law firms will move away from the billable hour toward alternative pricing such as fixed fees per transaction/fundraise and success fees in litigation.
“I think a lot of law firms are also experimenting with different pricing models where you do a fixed fee for a transaction or for a fundraise.”
WHO CLAIMED WHAT
Every position taken in the episode, attributed to its holder, with conviction (0–1) and the verbatim receipt.
Max Junestrand — 11 claims
opinionconviction 0.834:09
Legal services are only about 4% software and 96% services today, a mismatch that means the software share should be far larger and will grow into the service revenue.
“The software piece should be much bigger than that. And so the software piece naturally will grow into the service revenue”
assertionconviction 0.834:09
Legal is a supply-constrained market where demand for legal services vastly exceeds the available lawyers and services.
“The demand for legal services is much larger than what there are lawyers or legal services available.”
assertionconviction 0.835:16
The law-firm billable-hour model works by overcharging for associates and undercharging for partners, a structure AI directly undermines.
“you overcharge for the associates, and you actually undercharge for the partners.”
opinionconviction 0.843:25
Legacy legal-data incumbents will struggle to become AI-native because they can't attract the talent, don't work the hours, and are too politically internal to move quickly.
“They can't get the talent, they don't work our hours, and they're so political in their organizations that it's just hard to move.”
opinionconviction 0.7543:25
The early-AI bet that organizations owning the most data would be the winners is proving wrong in the market.
“I think at the outset of AI, many believed and made a bet that those organizations who had all the data was going to be the winners. As we're starting to see in the market, that's no longer the case.”
assertionconviction 0.8545:40
A viable legal-research product must contain all case data, not just the top 80% — the opposite of a power law — because top litigators need every case.
“You cannot build a legal research solution that doesn't have all of the data.”
opinionconviction 0.7541:12
Gathering every jurisdiction's cases, legislation, and regulatory updates at scale builds a real, durable data moat for Legora.
“But once you start to do that at scale, it builds a real data mode.”
opinionconviction 0.847:49
Anthropic's Claude legal offering is shallow (markdown skills plus a few integrations) and functions as a lead-generation funnel for Legora rather than real competition.
“you hit the ceiling or you understand how shallow it is, and then you call us. So it's actually a big pipeline generator for us.”
opinionconviction 0.8548:49
Fine-tuning or building a general legal-intelligence model is a total waste of time and money; only narrow models for narrow use cases make sense.
“So I don't believe in fine-tuning or building any general intelligence models. I think that's total waste of time and money. I do believe in very narrow models for narrow use cases”
assertionconviction 0.846:51
Following Opus 4.5 and 4.6, legal AI agents can now do real end-to-end case strategy, shifting the field from AI augmenting lawyers to AI doing the work while lawyers orchestrate.
“moving us from a world where AI is just augmenting, to AI is actually really doing things and your job becomes to orchestrate and to manage those agents as we're seeing in coding.”
opinionconviction 0.739:45
Junior-lawyer jobs will still exist, but the tasks shift from manual document review to orchestrating AI agents.
“The job will exist. The tasks will be different, right?”
Mati Stanislawski — 7 claims
opinionconviction 0.7525:52
ElevenLabs' model-agnostic platform is an advantage because customers can build their orchestration and voice layer without being locked to any single frontier model.
“And that being agnostic to a specific model is actually helpful because customers can make sure that they build a harness, build their agent orchestration, create a voice element of how that agent interacts with the world, how the marketing or interacts with the world, but they are not dependent on any model.”
opinionconviction 0.825:52
In voice models, the architecture matters more than scale — you have to fundamentally change how the model operates, not just make it bigger.
“It's the architecture that matters, not the scale. You really need to change how the model operates.”
assertionconviction 0.825:52
ElevenLabs has been able to out-compete Anthropic and OpenAI on voice models across text-to-speech, speech-to-text, turn-taking, and music.
“And we've been able to out-compete them on voice models, both on text to speech, speech to text, on the turn-taking, on music.”
assertionconviction 0.728:25
Frontier labs are actively trying to distill and use ElevenLabs' data, an ongoing threat ElevenLabs can only slow, not fully stop.
“We know that some companies are continuously trying to figure out how to distill and use the data. So, that is an existing problem and we have a few mechanisms to stop it. Or slow it down, not stop it.”
opinionconviction 0.728:25
ElevenLabs should own the interaction/communication layer rather than compete on knowledge work or coding.
“we won't focus on knowledge work, we won't focus on coding.”
opinionconviction 0.818:23
A person's voice is a form of identity and IP that others recognize and feel emotion from, making it both a risk and an opportunity.
“Yeah, the voice is identity and IP.”
assertionconviction 0.7514:34
People are more candid and honest with an AI voice agent than with a human because the emotional block (e.g., shame about a missed payment) is removed.
“With AI, people are much more open to share what actually happened, give the information, and suddenly this emotional block of like in front of other human, I don't want to be able to say all of that, is very different.”
Jason Calacanis — 1 claim
opinionconviction 0.6530:23
AGI has effectively already been achieved and simply hasn't been deployed yet.
“we're kind of there on AGI, don't you think? That we've kind of achieved it, we just haven't deployed it?”